Members and Ensembles

[bohaiboy]

Can someone please explain what is meant when models are referred to as ensembles or members? I was under the impression there were about 10 models, but a local private weatehr service keeps talking about members and ensembles.

TIA

[theavocado]

When a model is referred to, usually it is the high resolution, main set of calculations using the best starting data for it's initialization. This is also called the "deterministic" track in some circles. Some models also have a related ensemble comprised of many members. These are generally lower resolution such that they can all be run within the 6 or 12 hour cycle. With the ensemble, programmers "perturbate" the individual low-resolution runs with different starting inputs. This gives them an idea of the stability of the forecast given the current environmental conditions. If by making small changes all the members of the ensemble are tightly grouped, and are close to the deterministic track, the forecaster has a high confidence in the model for his/her forecast. If the spread of the ensemble members are spread out or don't match the deterministic track, the confidence is low.

If you watch The Price is Right, it can be compared to Plinko. I could seemingly drop each puck from the same spot, but it would land in a different slot each time. The tiny little perturbations of the weight of the chip, placement of the chip, way I let it go, etc. is enough to show that the system is unstable. Yet in bowling, a skilled player can easily replicate very similar lines for all thrown balls.